Case study
Live AI coaching across 40 sales calls a week
Operators were working from a script they had to remember while also listening. The system follows the conversation and feeds the next line, the objection handler, or the qualifying question to the operator while the call is still happening.
- Result
- ~40 calls/week
- Also
- Live, mid-conversation
- Client
- Sales operation
- Year
- 2026
The problem
Scripts fail in the moment. An operator who is listening properly cannot also scan a document, so the script gets abandoned early in the call and the quality gap between the best and worst operator stays wide.
What we built
- The system follows where the conversation has actually got to, rather than assuming it runs top to bottom.
- Objection handling surfaced the moment an objection appears, not afterwards in a review.
- Prompts short enough to read while talking, which is a real constraint and rules out most of what a model wants to write.
- Coverage across roughly 40 calls a week.
The result
- New operators work from live guidance rather than memory.
- The gap between the strongest and weakest operator narrows without more training time.
n8nClaudeMessaging APIs
Services used